[Mlir-commits] [mlir] [mlir][math] Fix the description of the semantics of math.clampf (PR #175012)

[llvmlistbot at llvm.org]

jack-slingsby wrote:

Thanks for all the discusion. I have updated the `math.clampf` description to be the original way round with `maxf(minf(...` but I have corrected the args. Overall this seems like the least invasive change both semantically and in terms of code changes. I have added a that if `min > max` the value is poison. I avoided mentioning any other spec and also avoided discussion about the alternative definition as imo this is bascially a spec itself and those points seems less appropriate with this in mind. The expansion pass is also updated. Comments are addressed :)


https://github.com/llvm/llvm-project/pull/175012